In order to solve the problem of transmission system monitoring in extreme weather, an intelligent monitoring system for transmission tower based on edge detection is proposed. Based on MobileNetV3 and multi-scale SSD, a fusion-level basic network model structure is designed, and an end-to-end ice thickness recognition model is established. The basic network is compressed based on discrimination-driven channel pruning, and the network can be deployed in monitoring terminals with limited computing resources. Taking the monitoring data provided by a power company as an example, the proposed model is verified. The results show that the mAP of the proposed model on the server side is 81.19%, and in the edge computation terminal, the discriminative perception based channel pruning method has a good effect on model compression, compared with the basic network, the mAP is reduced by about 3.67%, and the network scale is about 11 M.